Factor Analysis

Overview

This project implements factor analysis using the principal components (PC) and the iterated principal axis factoring (PAF) method with optional orthogonal rotations. The rotations are implemented via the factor_analyzer package.

Usage

First, instantiate a FactorAnalysis object

from factor_analysis import FactorAnalysis
fa = FactorAnalysis(n_factors=2, method="pc", rotation="varimax")

Then you can fit the model to a data set by executing

fa.fit(data)

and have a look the estimated parameters (e.g. loadings, communalities) with

fa.print_summary()

or by directly accessing the fitted attributes, which follow the naming convention used by the scikit-learn package (i.e. the method name ends with a _. For example:

fa.loadings_
fa.communalities_

To see which parameters the constructor accepts and which attributes are available, you can have a look at the FactorAnalysis class documentation.

The reproduced correlation matrix can be retrieved with the get_reprod_corr method.

Lastly, you can compute the factor scores with

fa.transform(data)

where data must be the same data used in the fit method.

You can also use the convenience method fit_transform to directly compute the factor scores after fitting.

scores = fa.fit_transform(data)

Dependencies

The required packages for the actual factor analysis class are

• scikit-learn
• numpy
• pandas
• factor_analyzer

In order to use the plotting functions in factor_analysis.plotting and to be able to execute the examples it is required to have

• jupyter
• matplotlib
• seaborn

installed.

To install the required packages, run (after activating a virtual environment)

pip install -r requirements.txt

If you want to run tests and commit to the project, then additionally run

pip install -r requirements-dev.txt
pre-commit install

Contributing

This project uses Poetry as its dependency manager. If you do not already have it, install Poetry. For instructions, follow this link:

https://github.com/python-poetry/poetry/blob/master/README.md#installation

Once you have Poetry installed, clone the repository and execute:

poetry config virtualenvs.in-project true
poetry install

This creates a virtual environment and installs the required packages.